On Apple, AI, and the loss of device-centric computing

I entered Apple’s world in 2006, through the unassuming portal of a 2005 iBook G4. My arrival was not guided by design or aesthetic allure, but by architecture. The machine behaved with a kind of mechanical honesty: predictable, legible, and true to its structure. Files dwelled in visible hierarchies. Processes could be summoned, inspected, and understood. Software ran in place, not elsewhere. The system invited study; an instrument whose response the practitioner could learn through discipline and familiarity.

In those early days, OS X treated the personal computer as the epicenter of computation. The network was an adjunct, a reach beyond, not a spine. Intelligence, broadly defined, resided within the silicon confines of the machine itself. I remained loyal through the transitions of architecture, the ascendance of iOS, and the continual reinvention of interfaces, because Apple’s core posture, technically if not philosophically, endured: it resisted the gravitational pull of the cloud. Syncing might touch remote systems, but the essential behaviors of the OS did not depend on them. Privacy was imperfect yet intrinsic, upheld by design rather than decree.

Now, with macOS 26 and iOS 26, that stance bends.

The fusion of Google’s Gemini into Siri is not a mere augmentation but a redirection of the current. It ushers an external, large-scale inference engine into the heart of the operating system’s intent interface. And Siri is no appendage; it is the bridge between thought and operation, the mesh tying reminders, messages, search, and automation together. When that bridge crosses into external computation, when meaning depends on a remote model, the device is no longer the seat of intelligence. It becomes a terminal.

Local inference keeps thought at home. Models run on the device’s own CPU, GPU, or neural cores; inputs, representations, and results never breach its borders. Performance is local, privacy structural, and failure graceful: if the network falters, the system simply knows its limits.

Remote inference is another kind of covenant. Inputs depart for distant computation; results return, often faster, but born of opacity. Behavior becomes contingent on connectivity, privacy mediated by policy, and latency the whim of distance. The system’s cognition becomes distributed, untraceable, and curiously non-local.

Once an operating system entrusts its core faculties to such mechanisms, it enshrines dependence within its design. Undoing that decision becomes not a revision, but a reformation. Apple may describe encryption boundaries, anonymized routing, and policy scaffolds, but these are trimmings. The architectural fact is stark: part of the OS’s mind now lives elsewhere.

This marks a migration, from device-centric computing toward service-centric orchestration. You can feel the same drift across Apple’s landscape. macOS and iOS increasingly assume the constant presence of iCloud. File hierarchies fade into abstractions. Sandboxing thickens. Services bud and multiply in the background. Once, the OS resembled a layered composition, its movements traceable. Now it behaves less like software and more like an environment managed from afar.

The signs are everywhere.

Spotlight, once a local index, now conjures blended results, a hybrid of home and elsewhere. Photos relies on off-device vision to recall faces and moments. Mail composes its own “smart” judgments, though from whose intelligence is uncertain. Shortcuts, once a deterministic language of automation, has absorbed the probabilistic hum of machine learning, bound now to the availability of remote inference.

Individually, these are conveniences. Together, they sketch a diagram of exodus: intelligence drifting outward, away from the device, into the networked mist.

This realignment mirrors Apple’s evolving business alignment. As revenue leans toward services, storage, and subscriptions, the OS becomes a delivery surface rather than a domain of exploration. Local computation turns optional, not assumed. The Gemini partnership brings this change into focus. Google’s role now extends beyond the app layer, reaching into the infrastructure itself.

Architecture, once chosen, compounds. When Siri can depend on remote reasoning, so can everything else. The more the system yields to off-device intelligence, the less intrinsic its own becomes. At length, iOS and Android, distinct in shape and polish, begin to share the same unseen substrate. Different shells. Common foundations.

Apple’s embrace of AI is hardly surprising. More striking is the suggestion that real intelligence belongs somewhere outside the machine. That direction reflects a choice, not a necessity, and it cuts against Apple’s own history of vertical integration, custom silicon, and tightly controlled ecosystems. The company is unusually well positioned to push intelligence closer to the device. Still, the pull of speed, parity, and scale appears to be winning out.

I used to admire Apple’s craft. Their devices remain exquisitely made, their systems refined to a rare degree. But my faith has thinned, not in quality, but in trajectory. Operating systems are the grammar of our digital thought. When their design bends toward clienthood, users begin to forget what independence feels like.

I hope for a re-centering: a reclaiming of locality. A world where devices host their own minds. Where the network assists, but never dictates. Where complexity remains comprehensible, and software feels once more like a structure one can inhabit and understand.

It’s a return to an older principle: a personal computer should be a place where computation happens, not just a place from which it’s requested.